Abstract: Ruminant production systems are important producers of food, support rural communities and culture, and help to maintain a range of ecosystem services including the sequestering of carbon in grassland soils. However, these systems also contribute significantly to climate change through greenhouse gas (GHG) emissions, while intensi- fication of production has driven biodiversity and nutrient loss, and soil degradation. Modeling can offer insights into the complexity underlying the relationships between climate change, management and policy choices, food production, and the maintenance of ecosystem services. This paper 1) provides an overview of how ruminant systems modeling supports the efforts of stakeholders and policymakers to predict, mitigate and adapt to climate change and 2) provides ideas for enhancing modeling to fulfil this role. Many grassland models can predict plant growth, yield and GHG emissions from mono-specific swards, but modeling multi-species swards, grassland quality and the impact of management changes requires further development. Current livestock models provide a good basis for predicting animal production; linking these with models of animal health and disease is a prior- ity. Farm-scale modeling provides tools for policymakers to predict the emissions of GHG and other pollutants from livestock farms, and to support the management decisions of farmers from environmental and economic standpoints. Other models focus on how policy and associated management changes affect a range of economic and environmental variables at regional, national and European scales. Models at larger scales generally utilise more empirical approaches than those applied at animal, field and farm-scales and include assumptions which may not be valid under climate change conditions. It is therefore important to continue to develop more realistic representations of processes in regional and global models, using the understanding gained from finer-scale modeling. An iterative process of model development, in which lessons learnt from mechanistic models are ap- plied to develop ‘smart’ empirical modeling, may overcome the trade-off between complexity and usability. De- veloping the modeling capacity to tackle the complex challenges related to climate change, is reliant on closer links between modelers and experimental researchers, and also requires knowledge-sharing and increasing technical compatibility across modeling disciplines. Stakeholder engagement throughout the process of model development and application is vital for the creation of relevant models, and important in reducing problems re- lated to the interpretation of modeling outcomes. Enabling modeling to meet the demands of policymakers and other stakeholders under climate change will require collaboration within adequately-resourced, long-term inter-disciplinary research networks

Abstract: Process-based grassland models (PBMs) simulate growth and development of vegetation over time. The models tend to have a large number of parameters that represent properties of the plants. To simulate different cultivars of the same species, different parameter values are required. Parameter differences may be interpreted as genetic variation for plant traits. Despite this natural connection between PBMs and plant genetics, there are only few examples of successful use of PBMs in plant breeding. Here we present a new procedure by which PBMs can help design ideotypes, i.e. virtual cultivars that optimally combine properties of existing cultivars. Ideotypes constitute selection targets for breeding. The procedure consists of four steps: (1) Bayesian calibration of model parameters using data from cultivar trials, (2) Estimating genetic variation for parameters from the combination of cultivar-specific calibrated parameter distributions, (3) Identifying parameter combinations that meet breeding objectives, (4) Translating model results to practice, i.e. interpreting parameters in terms of practical selection criteria. We show an application of the procedure to timothy (Phleum pratense L.) as grown in different regions of Norway.

Abstract: Highlights • The parameters to be fixed were consistent across sites. • Model calibration must be performed separately for each specific case. • Possible to reduce model parameters from 66 to 45. • Strong model reductions must be avoided. • The error term for the training data were characterised by timing (phase shift). Abstract Proper parameterisation and quantification of model uncertainty are two essential tasks in improvement and assessment of model performance. Bayesian calibration is a method that combines both tasks by quantifying probability distributions for model parameters and outputs. However, the method is rarely applied to complex models because of its high computational demand when used with high-dimensional parameter spaces. We therefore combined Bayesian calibration with sensitivity analysis, using the screening method by Morris (1991), in order to reduce model complexity by fixing parameters to which model output was only weakly sensitive to a nominal value. Further, the robustness of the model with respect to reduction in the number of free parameters were examined according to model discrepancy and output uncertainty. The process-based grassland model BASGRA was examined in the present study on two sites in Norway and in Germany, for two grass species (Phleum pratense and Arrhenatherum elatius). According to this study, a reduction of free model parameters from 66 to 45 was possible. The sensitivity analysis showed that the parameters to be fixed were consistent across sites (which differed in climate and soil conditions), while model calibration had to be performed separately for each combination of site and species. The output uncertainty decreased slightly, but still covered the field observations of aboveground biomass. Considering the training data, the mean square error for both the 66 and the 45 parameter model was dominated by errors in timing (phase shift), whereas no general pattern was found in errors when using the validation data. Stronger model reduction should be avoided, as the error term increased and output uncertainty was underestimated.